BestGiniOfColumn< LineSearchLossTag > Class Template Reference

#include <vigra/random_forest/rf_split.hxx>

Public Member Functions
template<class DataSourceF_t , class DataSource_t , class I_Iter , class Array >
void	operator() (DataSourceF_t const &column, DataSource_t const &labels, I_Iter &begin, I_Iter &end, Array const &region_response)

Detailed Description

template<class LineSearchLossTag>
class vigra::BestGiniOfColumn< LineSearchLossTag >

Given a column, choose a split that minimizes some loss

Member Function Documentation

void operator()	(	DataSourceF_t const &	column,
		DataSource_t const &	labels,
		I_Iter &	begin,
		I_Iter &	end,
		Array const &	region_response
	)

calculate the best gini split along a Feature Column

Parameters

column	the feature vector - has to support the [] operator
labels	the label vector
begin
end	(in and out) begin and end iterators to the indices of the samples in the current region. the range begin - end is sorted by the column supplied during function execution.
region_response	??? class histogram of the range.

precondition: begin, end valid range, class_counts positive integer valued array with the class counts in the current range. labels.size() >= max(begin, end); postcondition: begin, end sorted by column given. min_gini_ contains the minimum gini found or NumericTraits<double>::max if no split was found. min_index_ contains the splitting index in the range or invalid data if no split was found. BestCirremtcounts[0] and [1] contain the class histogram of the left and right region of the left and right regions.

---

The documentation for this class was generated from the following file:

• vigra/random_forest/rf_split.hxx